@whuber thanks for that little snippet of R code on the Tukey resistant line; it gives a clear suggestion about the kind of problem. Anyone know - how does one report an R-bug properly? This would seem a good time to also suggest putting in a CI and test for the slope (e.g. see p12-15 here). Would a post to the R-devel mailing list be the right place to take this?

The C code for the function is here : svn.r-project.org/R/trunk/src/library/stats/src/line.c ... I'm now trying to get my brain to go back a couple of decades and remember how to read C code... it's been about that long.